9  Trigger Configuration: HND & SLV

9.1 Overview

?sec-honduras-trigger developed a purely rainfall-based trigger optimization for Honduras, defining drought as the N driest ERA5 years (matched return period). This chapter takes a different approach: we begin with humanitarian impact data to identify which years must be triggered, then layer in ERA5 rainfall skill to select the final configuration

The key insight motivating this approach is that seasonal rainfall forecasts are an imperfect proxy for drought impact. EM-DAT records and CERF allocations provide direct evidence of when humanitarian consequences occurred. Rather than optimizing solely against observed rainfall, we first ensure our trigger catches the most consequential events, then optimize for rainfall prediction accuracy among configurations that meet that constraint.

This chapter covers Honduras, El Salvador, and Guatemala. For Guatemala, we evaluate two separate areas of interest: the standard OCHA AOI (Chiquimula and Alta Verapaz) and the StartNetwork AOI (Quiché and Baja Verapaz).

9.2 Impact Data Landscape

Before defining trigger thresholds, we examine the historical record of drought impacts across the dry corridor. From this evidence we define two priority classes for the trigger optimization:

  • Priority 1 (must-hit): The CERF hypothesis years — 2009, 2014, 2015, 2018. These are the years where CERF actually allocated drought funding (adjusted for delayed response timing). They represent donor-validated humanitarian crises where we have the strongest obligation to trigger.

  • Priority 2 (should-hit): Years where all 3 countries appear in EM-DAT expanded drought records but are not CERF hypothesis years — 1994, 2001, 2019. Regional consensus across all three countries indicates widespread drought conditions, even without CERF activation.

Table 9.1 shows the full landscape of CERF allocations, EM-DAT entries (both as-reported and expanded across multi-year events), and the resulting priority classification for 1991–2024.

Table 9.1: Impact data summary across CERF, EM-DAT, and CERF hypothesis years. Priority 1 (pink) = CERF hypothesis drought years. Priority 2 (yellow) = years where all 3 countries appear in EM-DAT expanded.
Impact Data Summary (CERF, EM-DAT, CERF Hypothesis)
1991–2024 | Priority 1 = CERF hypothesis | Priority 2 = 3/3 countries EM-DAT expanded
Year CERF EM-DAT (reported) EM-DAT (expanded) Priority
HND GTM SLV Hypothesis HND GTM SLV HND GTM SLV
1991
1992
1993
1994 ✓ ✓ ✓ ✓ ✓ ✓ 2
1995
1996
1997 ✓ ✓
1998 ✓ ✓ ✓
1999
2000 ✓ ✓
2001 ✓ ✓ ✓ ✓ ✓ ✓ 2
2002 ✓ ✓
2003 ✓
2004 ✓ ✓
2005
2006
2007
2008
2009 ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ 1
2010 ✓ ✓
2011 ✓
2012 ✓ ✓ ✓ ✓
2013
2014 ✓ ✓ ✓ ✓ ✓ ✓ ✓ 1
2015 ✓ ✓ ✓ ✓ ✓ ✓ ✓ 1
2016 ✓ ✓ ✓
2017
2018 ✓ ✓ ✓ ✓ ✓ ✓ ✓ 1
2019 ✓ ✓ ✓ ✓ ✓ 2
2020
2021
2022
2023 ✓ ✓
2024

9.3 Rainfall vs Impact Alignment

A key question is whether seasonal rainfall alone can predict humanitarian impact. The tables below show per-country validation: for each year, we compare ERA5 observed rainfall ranking with CERF allocations, EM-DAT records, and total affected population.

The validation tables below cross-reference each year’s primera and postrera rainfall dry rank with CERF allocations, EM-DAT records, and total affected population. Years are sorted by maximum seasonal dryness rank so the most extreme rainfall years appear first. A mismatch between dry rank and impact columns indicates that rainfall alone is not a reliable predictor of humanitarian consequences.

Table 9.2
Impact Validation: HND (1991–2024)
Sorted by max_rp (desc)
Year CERF EM-DAT Affected ERA5 Obs RP
Country Regional Hypothesis Reported Interpolated Pri Post Max
1994
0
0
35.0 3.2 35.0
2012
125K
125K
1.3 35.0 35.0
1992 5.0 17.5 17.5
2001
195K
49K
17.5 5.0 17.5
2004
138K
186K
11.7 4.4 11.7
2021 3.5 11.7 11.7
2000
1K
1K
8.8 3.9 8.8
2009 ✓ ✓ ✓
45K
15K
1.8 8.8 8.8
1991 7.0 5.8 7.0
2002
220K
269K
1.6 7.0 7.0
2019 ✓
180K
5.8 1.9 5.8
2015 ✓ ✓ ✓
1.0M
4.4 1.5 4.4
2014 ✓ ✓ ✓
2.0M
1.0M
3.9 1.4 3.9
2003
49K
1.9 3.5 3.5
2013 3.2 2.2 3.2
1997
0
0
2.9 2.1 2.9
2010 ✓ ✓
15K
1.0 2.9 2.9
2016 ✓ 2.5 2.7 2.7
2023
10K
10K
2.7 1.3 2.7
1993 1.1 2.5 2.5
2006 2.1 2.3 2.3
2018 ✓
360K
180K
2.3 1.8 2.3
1998
0
2.2 1.0 2.2
1999 1.8 1.1 1.8
2005 1.1 1.8 1.8
2007 1.7 1.3 1.7
2011
15K
1.1 1.7 1.7
1995 1.2 1.6 1.6
2008 1.5 1.2 1.5
1996 1.5 1.5 1.5
2020 1.4 1.1 1.4
2024 1.3 1.1 1.3
2017 1.2 1.2 1.2
2022 1.2 1.2 1.2
Table 9.3
Impact Validation: SLV (1991–2024)
Sorted by max_rp (desc)
Year CERF EM-DAT Affected ERA5 Obs RP
Country Regional Hypothesis Reported Interpolated Pri Post Max
2012 3.2 35.0 35.0
2015 ✓ ✓ ✓
700K
350K
35.0 1.2 35.0
1994
0
0
17.5 4.4 17.5
2021 1.5 17.5 17.5
1997 11.7 2.1 11.7
2001
400K
400K
7.0 11.7 11.7
1991 5.8 8.8 8.8
2018 ✓
387K
193K
8.8 1.5 8.8
2016 ✓
350K
4.4 7.0 7.0
2009 ✓ ✓
0
0
2.3 5.8 5.8
1992 2.1 5.0 5.0
2023 5.0 2.2 5.0
2003 1.9 3.9 3.9
2014 ✓ ✓ 3.9 1.8 3.9
2000 3.5 2.7 3.5
2002 2.7 3.5 3.5
2010 ✓ 1.0 3.2 3.2
2013 2.5 2.9 2.9
2019 ✓ ✓
193K
2.9 1.6 2.9
2006 1.8 2.5 2.5
2004 2.2 2.3 2.3
1993 1.5 1.9 1.9
1998
0
0
1.8 1.0 1.8
1996 1.2 1.8 1.8
2007 1.7 1.5 1.7
2017 1.4 1.7 1.7
1999 1.6 1.1 1.6
2024 1.1 1.4 1.4
2005 1.1 1.3 1.3
2011 1.3 1.2 1.3
1995 1.1 1.3 1.3
2022 1.3 1.2 1.3
2020 1.2 1.1 1.2
2008 1.2 1.1 1.2
Table 9.4
Impact Validation: GTM (1991–2024)
Sorted by max_rp (desc)
Year CERF EM-DAT Affected ERA5 Obs RP
Country Regional Hypothesis Reported Interpolated Pri Post Max
2019 ✓ ✓
750K
35.0 2.2 35.0
2021 1.8 35.0 35.0
1994
0
0
17.5 5.8 17.5
2012
266K
266K
1.3 17.5 17.5
2004 2.3 11.7 11.7
2015 ✓ ✓
433K
11.7 1.2 11.7
2010 ✓ 1.0 8.8 8.8
2023 8.8 1.6 8.8
2009 ✓ ✓ ✓
2.5M
2.5M
2.7 7.0 7.0
2018 ✓
1.5M
750K
7.0 5.0 7.0
1998 5.8 1.1 5.8
1991 5.0 4.4 5.0
2002 4.4 3.5 4.4
1992 1.5 3.9 3.9
2016 ✓ ✓
433K
3.9 2.7 3.9
2001
114K
114K
3.5 2.9 3.5
1993 1.2 3.2 3.2
2003 3.2 1.9 3.2
1997 2.9 1.5 2.9
2005 1.1 2.5 2.5
2007 2.5 1.8 2.5
2006 1.1 2.3 2.3
2024 2.2 1.3 2.2
1999 2.1 1.2 2.1
2000 1.5 2.1 2.1
2011 1.9 1.1 1.9
1995 1.4 1.8 1.8
2014 ✓ ✓ ✓
1.3M
433K
1.8 1.2 1.8
1996 1.3 1.7 1.7
2008 1.7 1.1 1.7
2022 1.6 1.3 1.6
2017 1.2 1.5 1.5
2013 1.1 1.4 1.4
2020 1.2 1.0 1.2
Table 9.5
Impact Validation: GTM (1991–2024)
Sorted by max_rp (desc)
Year CERF EM-DAT Affected ERA5 Obs RP
Country Regional Hypothesis Reported Interpolated Pri Post Max
1994
0
0
35.0 7.0 35.0
2021 1.5 35.0 35.0
2012
266K
266K
1.1 17.5 17.5
2019 ✓ ✓
750K
17.5 1.9 17.5
2004 11.7 5.0 11.7
2009 ✓ ✓ ✓
2.5M
2.5M
2.5 11.7 11.7
1991 5.0 8.8 8.8
2015 ✓ ✓
433K
8.8 1.5 8.8
2002 7.0 2.5 7.0
2016 ✓ ✓
433K
3.5 5.8 5.8
2023 5.8 1.7 5.8
2018 ✓
1.5M
750K
4.4 4.4 4.4
2003 3.9 2.7 3.9
2010 ✓ 1.0 3.9 3.9
1993 1.3 3.5 3.5
1992 1.8 3.2 3.2
2007 3.2 1.6 3.2
1998 2.9 1.0 2.9
2006 1.1 2.9 2.9
1997 2.7 1.3 2.7
2001
114K
114K
2.3 2.3 2.3
1996 1.3 2.2 2.2
1999 2.2 1.3 2.2
2005 1.2 2.1 2.1
2014 ✓ ✓ ✓
1.3M
433K
2.1 1.2 2.1
2022 1.9 1.2 1.9
2017 1.2 1.8 1.8
2024 1.8 1.5 1.8
2000 1.6 1.8 1.8
2011 1.7 1.1 1.7
1995 1.5 1.4 1.5
2008 1.4 1.1 1.4
2020 1.2 1.1 1.2
2013 1.1 1.2 1.2

These tables demonstrate a recurring pattern: some of the most impactful drought years (e.g. those with CERF allocations or high total affected) do not always coincide with the driest rainfall years, and vice versa. This imperfect correlation between rainfall and impact is the primary motivation for our hybrid optimization approach.

9.4 Optimization Strategy

Given that rainfall is an imperfect predictor of humanitarian impact, we adopt a staged optimization that prioritizes impact evidence first, then maximizes rainfall prediction skill:

  1. Impact-first filtering: Evaluate all threshold configurations against EM-DAT impact years using standard F1. Keep only configs that trigger in all Priority 1 years (must-hit). Among those, keep configs with the maximum Priority 2 hits (should-hit).

  2. ERA5 rainfall evaluation: For the surviving candidate pool, evaluate against ERA5 observed rainfall using the tolerant matched-RP F1 from ?sec-honduras-trigger. This tells us how well each config predicts actual rainfall anomalies, independent of the impact definition.

  3. Combined selection: Plot impact F1 against ERA5 tolerant F1 to visualize the tradeoff. Compute an average of both F1 scores and identify the best-performing config at each return period level. Among tied configs, select the one with the lowest threshold coefficient of variation (most operationally uniform thresholds).

The rationale is straightforward: since SEAS5 forecasts are our only trigger variable, and they predict rainfall rather than impact directly, it makes sense to first ensure we capture the most consequential events (via impact filtering), then optimize for what the forecast actually predicts well (rainfall). After the priority-hit filters are satisfied, a high ERA5 tolerant F1 indicates the trigger is selecting genuinely dry years rather than succeeding on impact years by coincidence.

9.5 Country-Specific Results

Honduras uses departments El Paraiso (HN07) and Francisco Morazan (HN08) as the area of interest, with 3 primera leadtimes and 3 postrera leadtimes. EM-DAT records 17 impact years in 1991–2024, giving an empirical impact RP of 3.0 years.

Impact Evaluation & Priority Filtering

After generating 8,526 constrained configurations and evaluating against impact years, priority filtering retains 2,613 candidates that hit 100% of Priority 1 years and 67% of Priority 2 years.

ERA5 Evaluation & Combined Scoring

The plot below shows each candidate configuration positioned by its ERA5 tolerant F1 (x-axis) and impact F1 (y-axis). Configurations near the top-right corner perform well on both metrics. Circled points mark selected options per RP level: A = best average F1, B = best ERA5 tolerant F1. Yellow halos mark fully uniform configs (all thresholds equal across all leadtimes and seasons) — an operationally favourable baseline that simplifies trigger communication. Uniform configs are drawn from the pre-filter pool (before must-hit/should-hit filtering) so stricter options that wouldn’t pass priority filters are still visible as reference points. Hover over any point for details; click circled or haloed points to view the full per-leadtime activation table.

Figure 9.1: ERA5 tolerant F1 vs Impact F1 for Honduras. Click a circled point to see its trigger activation detail below.
Table 9.6
Honduras: Configuration Summary
# Config Annual RP Primera RP Postrera RP Primera Thresholds Postrera Thresholds
1 RP 2.7 A 2.69 3.50 3.89 LT0: 4.4 | LT1: 5.8 | LT2: 7.0 LT0: 4.4 | LT1: 5.8 | LT2: 7.0
2 RP 2.7 (Uniform thresh 5.0) 2.69 3.18 3.89 LT0: 5.0 | LT1: 5.0 | LT2: 5.0 LT0: 5.0 | LT1: 5.0 | LT2: 5.0
3 RP 2.9 B 2.92 3.50 3.89 LT0: 5.0 | LT1: 5.8 | LT2: 7.0 LT0: 5.8 | LT1: 5.8 | LT2: 5.8
4 RP 3.2 B 3.18 3.50 5.83 LT0: 5.0 | LT1: 5.8 | LT2: 7.0 LT0: 8.8 | LT1: 8.8 | LT2: 8.8
5 RP 2.5 (Uniform thresh 4.4) 2.50 3.18 3.50 LT0: 4.4 | LT1: 4.4 | LT2: 4.4 LT0: 4.4 | LT1: 4.4 | LT2: 4.4
6 RP 2.7 (Uniform thresh 5.8) 2.69 3.50 3.89 LT0: 5.8 | LT1: 5.8 | LT2: 5.8 LT0: 5.8 | LT1: 5.8 | LT2: 5.8
7 RP 2.9 A 2.92 3.18 7.00 LT0: 5.0 | LT1: 5.8 | LT2: 5.8 LT0: 8.8 | LT1: 11.7 | LT2: 11.7
8 RP 3.2 A 3.18 3.50 7.00 LT0: 5.0 | LT1: 5.8 | LT2: 8.8 LT0: 8.8 | LT1: 11.7 | LT2: 11.7
9 RP 3.5 A 3.50 3.50 8.75 LT0: 5.0 | LT1: 5.8 | LT2: 17.5 LT0: 17.5 | LT1: 17.5 | LT2: 17.5
10 RP 3.5 (Uniform thresh 7.0) 3.50 4.38 5.00 LT0: 7.0 | LT1: 7.0 | LT2: 7.0 LT0: 7.0 | LT1: 7.0 | LT2: 7.0
11 RP 3.5 (Uniform thresh 8.8) 3.50 4.38 5.83 LT0: 8.8 | LT1: 8.8 | LT2: 8.8 LT0: 8.8 | LT1: 8.8 | LT2: 8.8
12 RP 3.9 (Uniform thresh 11.7) 3.89 5.00 7.00 LT0: 11.7 | LT1: 11.7 | LT2: 11.7 LT0: 11.7 | LT1: 11.7 | LT2: 11.7
13 RP 5.8 (Uniform thresh 17.5) 5.83 7.00 8.75 LT0: 17.5 | LT1: 17.5 | LT2: 17.5 LT0: 17.5 | LT1: 17.5 | LT2: 17.5

Trigger Activation Detail (per leadtime)

Click a circled point on the tradeoff plot above to swap the detailed per-leadtime activation table below.

Table 9.7
Trigger Activation: HND (1991–2024)
Green = trigger fires at that leadtime | Annual RP = 2.92 (RP 2.9 A)
Year P CERF EM-DAT Affected Primera Forecast RP (3.2) Postrera Forecast RP (7.0) ERA5 Obs RP
Reported Interpolated LT0 (May) RP 5.0 LT1 (Apr) RP 5.8 LT2 (Mar) RP 5.8 LT0 (Sep) RP 8.8 LT1 (Aug) RP 11.7 LT2 (Jul) RP 11.7 Pri Post
1991
35.0 3.9 7.0 8.8 17.5 17.5 7.0 5.8
1992
1.6 1.5 1.2 11.7 3.9 11.7 5.0 17.5
1994 2
0
0
7.0 4.4 17.5 7.0 7.0 4.4 35.0 3.2
1997
0
0
5.0 2.5 2.9 5.8 35.0 8.8 2.9 2.1
2006
2.2 1.8 5.8 1.9 1.7 1.4 2.1 2.3
2009 1 ✓
45K
15K
4.4 35.0 35.0 17.5 4.4 3.9 1.8 8.8
2014 1 ✓
2.0M
1.0M
3.2 5.8 4.4 3.5 5.0 5.8 3.9 1.4
2015 1 ✓
1.0M
17.5 8.8 11.7 35.0 11.7 35.0 4.4 1.5
2018 1
360K
180K
8.8 7.0 3.9 5.0 8.8 7.0 2.3 1.8
2019 2
180K
5.8 17.5 5.0 2.2 2.2 3.2 5.8 1.9
2021
11.7 5.0 3.2 2.5 2.5 1.8 3.5 11.7
2023
10K
10K
1.9 11.7 8.8 2.7 2.9 2.1 2.7 1.3
1993
1.2 1.5 1.8 3.9 3.5 2.9 1.1 2.5
1995
1.8 1.4 1.5 1.1 1.1 1.3 1.2 1.6
1996
1.2 1.6 1.6 1.8 2.1 1.3 1.5 1.5
1998
0
1.4 1.2 1.0 1.1 1.2 1.2 2.2 1.0
1999
1.5 1.2 1.9 1.1 1.1 1.1 1.8 1.1
2000
1K
1K
2.1 2.2 1.7 1.8 1.8 1.7 8.8 3.9
2001 2
195K
49K
2.5 3.5 2.5 3.2 2.3 2.3 17.5 5.0
2002
220K
269K
3.9 1.7 1.3 2.9 5.8 5.0 1.6 7.0
2003
49K
1.8 2.1 1.5 2.1 1.6 1.9 1.9 3.5
2004
138K
186K
1.3 1.2 1.3 2.3 2.7 2.7 11.7 4.4
2005
1.1 1.1 1.1 1.5 1.2 1.1 1.1 1.8
2007
1.1 1.1 1.2 1.3 1.5 1.6 1.7 1.3
2008
2.9 2.9 2.3 1.6 1.2 1.2 1.5 1.2
2010
✓
15K
1.0 1.0 1.1 1.0 1.0 1.0 1.0 2.9
2011
15K
1.1 1.9 2.1 1.2 1.1 1.1 1.1 1.7
2012
125K
125K
2.7 2.7 3.5 4.4 3.2 1.5 1.3 35.0
2013
1.3 1.3 1.4 1.4 1.5 2.5 3.2 2.2
2016
1.5 1.3 1.2 1.5 1.8 3.5 2.5 2.7
2017
1.7 2.3 1.8 1.7 1.9 2.2 1.2 1.2
2020
3.5 1.8 2.7 1.2 1.3 1.8 1.4 1.1
2022
2.3 3.2 2.2 1.2 1.3 1.5 1.2 1.2
2024
1.2 1.1 1.1 1.3 1.4 1.2 1.3 1.1

Activation Summary

Table 9.8
Honduras: Activation Summary
Pr = primera only | Pst = postrera only | Pr+Pst = both seasons
Year P CERF EM-DAT ERA5 Obs RP Activation
Pri Post RP 2.7 A RP 2.9 A RP 2.9 B RP 3.2 A RP 3.2 B RP 3.5 A RP 2.5 (Uniform thresh 4.4) RP 2.7 (Uniform thresh 5.0) RP 2.7 (Uniform thresh 5.8) RP 3.5 (Uniform thresh 7.0) RP 3.5 (Uniform thresh 8.8) RP 3.9 (Uniform thresh 11.7) RP 5.8 (Uniform thresh 17.5)
1991 7.0 5.8 Pr+Pst Pr+Pst Pr+Pst Pr+Pst Pr+Pst Pr+Pst Pr+Pst Pr+Pst Pr+Pst Pr+Pst Pr+Pst Pr+Pst Pr+Pst
1992 5.0 17.5 Pst Pst Pst Pst Pst Pst Pst Pst Pst Pst Pst
1994 P2
0
35.0 3.2 Pr+Pst Pr Pr+Pst Pr Pr Pr Pr+Pst Pr+Pst Pr+Pst Pr+Pst Pr Pr Pr
1997
0
2.9 2.1 Pr+Pst Pr+Pst Pr+Pst Pr+Pst Pr+Pst Pr+Pst Pr+Pst Pr+Pst Pst Pst Pst Pst Pst
2002
269K
1.6 7.0 Pst Pst Pst Pst Pst
2006 2.1 2.3 Pr Pr Pr Pr
2009 P1 ✓
15K
1.8 8.8 Pr+Pst Pr+Pst Pr+Pst Pr+Pst Pr+Pst Pr+Pst Pr+Pst Pr+Pst Pr+Pst Pr+Pst Pr+Pst Pr+Pst Pr+Pst
2012
125K
1.3 35.0 Pst Pst
2014 P1 ✓
1.0M
3.9 1.4 Pr Pr Pr+Pst Pr Pr Pr Pr+Pst Pr+Pst Pr+Pst
2015 P1 ✓
1.0M
4.4 1.5 Pr+Pst Pr+Pst Pr+Pst Pr+Pst Pr+Pst Pr+Pst Pr+Pst Pr+Pst Pr+Pst Pr+Pst Pr+Pst Pr+Pst Pr+Pst
2018 P1
180K
2.3 1.8 Pr+Pst Pr Pr+Pst Pr Pr+Pst Pr Pr+Pst Pr+Pst Pr+Pst Pr+Pst Pr+Pst
2019 P2
180K
5.8 1.9 Pr Pr Pr Pr Pr Pr Pr Pr Pr Pr Pr Pr Pr
2021 3.5 11.7 Pr Pr Pr Pr Pr Pr Pr Pr Pr Pr Pr Pr
2023
10K
2.7 1.3 Pr Pr Pr Pr Pr Pr Pr Pr Pr Pr Pr Pr
1993 1.1 2.5
1995 1.2 1.6
1996 1.5 1.5
1998
0
2.2 1.0
1999 1.8 1.1
2000
1K
8.8 3.9
2001 P2
49K
17.5 5.0
2003
49K
1.9 3.5
2004
186K
11.7 4.4
2005 1.1 1.8
2007 1.7 1.3
2008 1.5 1.2
2010 ✓
15K
1.0 2.9
2011
15K
1.1 1.7
2013 3.2 2.2
2016 2.5 2.7
2017 1.2 1.2
2020 1.4 1.1
2022 1.2 1.2
2024 1.3 1.1
Table 9.9
Honduras: Topline Performance
34-year record (1991–2024) | 17 EM-DAT impact years
Config Annual RP Priority Years1 vs EM-DAT Impact2 vs ERA5 Rainfall3 Years Fired
P1 Caught P2 Caught TP FP FN TN F1 TP FP FN TN F1
RP 2.7 A 2.69 4/4 2/3 10 3 7 14 0.667 11 (14) 8 (5) 8 41 0.679 1991 1992 1994 1997 2002 2009 2012 2014 2015 2018 2019 2021 2023
RP 2.9 B 2.92 4/4 2/3 9 3 8 14 0.621 10 (13) 9 (6) 9 40 0.627 1991 1992 1994 1997 2002 2009 2014 2015 2018 2019 2021 2023
RP 3.2 B 3.18 4/4 2/3 8 3 9 14 0.571 9 (11) 7 (5) 7 45 0.614 1991 1992 1994 1997 2009 2014 2015 2018 2019 2021 2023
RP 2.7 (Uniform thresh 5.0) 2.69 4/4 2/3 9 4 8 13 0.600 10 (14) 10 (6) 10 38 0.623 1991 1992 1994 1997 2002 2006 2009 2014 2015 2018 2019 2021 2023
Primera Both Postrera Red outline = not all configs fire
1 P1 = CERF-funded drought years (highest confidence). P2 = all 3 countries in EM-DAT (regional consensus).
2 TP = trigger fires in an impact year. FP = fires in a non-impact year (may still be genuinely dry). FN = misses an impact year.
3 ERA5: summed across primera + postrera. TP & FP shown as matched (tolerant). FN = matched only (tolerant FN equals matched by design). Tolerant = widened drought definition (RP − buffer). F1 computed from tolerant TP, tolerant FP, matched FN.
Table 9.10: Honduras recommended configurations: best average F1 per return period, tiebroken by lowest cross-season threshold CV.
Honduras: Recommended Configurations
A = max avg F1 | B = max ERA5 tol F1 | U = uniform (all thresholds equal)
Return periods Return periods
config_label Postrera thresholds (RP) Primera thresholds (RP) annual_rp Performance
Criterion s0 s1 s2 s_seasonal_rp p0 p1 p2 p_seasonal_rp impact_f1 era5_tol_f1 avg_f1 mean_cv all_cv
RP 2.7 A both 4.375000 5.833333 7.000000 3.89 4.375000 5.833333 7.000000 3.50 2.69 0.566 0.679 0.622 0.229 0.205
RP 2.9 A max ERA5 tol F1 8.750000 11.666667 11.666667 7.00 5.000000 5.833333 5.833333 3.18 2.92 0.422 0.633 0.527 0.122 0.372
RP 2.9 B max avg F1 5.833333 5.833333 5.833333 3.89 5.000000 5.833333 7.000000 3.50 2.92 0.566 0.627 0.596 0.085 0.108
RP 3.2 A max ERA5 tol F1 8.750000 11.666667 11.666667 7.00 5.000000 5.833333 8.750000 3.50 3.18 0.433 0.636 0.535 0.230 0.326
RP 3.2 B max avg F1 8.750000 8.750000 8.750000 5.83 5.000000 5.833333 7.000000 3.50 3.18 0.470 0.614 0.542 0.085 0.226
RP 3.5 A both 17.500000 17.500000 17.500000 8.75 5.000000 5.833333 17.500000 3.50 3.50 0.439 0.489 0.464 0.370 0.464
RP 2.5 (Uniform thresh 4.4) uniform 4.375000 4.375000 4.375000 3.50 4.375000 4.375000 4.375000 3.18 2.50 0.582 0.646 0.614 0.000 0.000
RP 2.7 (Uniform thresh 5.0) uniform 5.000000 5.000000 5.000000 3.89 5.000000 5.000000 5.000000 3.18 2.69 0.555 0.623 0.589 0.000 0.000
RP 2.7 (Uniform thresh 5.8) uniform 5.833333 5.833333 5.833333 3.89 5.833333 5.833333 5.833333 3.50 2.69 0.528 0.596 0.562 0.000 0.000
RP 3.5 (Uniform thresh 7.0) uniform 7.000000 7.000000 7.000000 5.00 7.000000 7.000000 7.000000 4.38 3.50 0.448 0.448 0.448 0.000 0.000
RP 3.5 (Uniform thresh 8.8) uniform 8.750000 8.750000 8.750000 5.83 8.750000 8.750000 8.750000 4.38 3.50 0.414 0.414 0.414 0.000 0.000
RP 3.9 (Uniform thresh 11.7) uniform 11.666667 11.666667 11.666667 7.00 11.666667 11.666667 11.666667 5.00 3.89 0.345 0.345 0.345 0.000 0.000
RP 5.8 (Uniform thresh 17.5) uniform 17.500000 17.500000 17.500000 8.75 17.500000 17.500000 17.500000 7.00 5.83 0.325 0.325 0.325 0.000 0.000

El Salvador uses country-level (admin 0) aggregation with 3 primera leadtimes and 2 postrera leadtimes. EM-DAT records 8 impact years. We use a target RP of 3.

Impact Evaluation & Priority Filtering

After priority filtering, 1,381 candidates remain, hitting 100% of Priority 1 years and 67% of Priority 2 years.

ERA5 Evaluation & Combined Scoring

Figure 9.2: ERA5 tolerant F1 vs Impact F1 for El Salvador. Click a circled point to see its trigger activation detail below.
Table 9.11
El Salvador: Configuration Summary
# Config Annual RP Primera RP Postrera RP Primera Thresholds Postrera Thresholds
1 RP 2.9 A (uniform) 2.92 3.89 3.89 LT0: 5.0 | LT1: 5.0 | LT2: 5.0 LT0: 5.0 | LT1: 5.0
2 RP 2.7 A (uniform) 2.69 3.50 3.89 LT0: 4.4 | LT1: 4.4 | LT2: 4.4 LT0: 4.4 | LT1: 4.4
3 RP 3.2 A 3.18 3.89 4.38 LT0: 5.0 | LT1: 5.0 | LT2: 5.8 LT0: 5.0 | LT1: 7.0
4 RP 3.5 A 3.50 3.89 5.00 LT0: 5.0 | LT1: 5.0 | LT2: 5.8 LT0: 5.8 | LT1: 7.0
5 RP 3.5 (Uniform thresh 5.8) 3.50 4.38 4.38 LT0: 5.8 | LT1: 5.8 | LT2: 5.8 LT0: 5.8 | LT1: 5.8
6 RP 4.4 (Uniform thresh 7.0) 4.38 4.38 5.83 LT0: 7.0 | LT1: 7.0 | LT2: 7.0 LT0: 7.0 | LT1: 7.0
7 RP 4.4 (Uniform thresh 8.8) 4.38 5.00 5.83 LT0: 8.8 | LT1: 8.8 | LT2: 8.8 LT0: 8.8 | LT1: 8.8
8 RP 5.0 (Uniform thresh 11.7) 5.00 5.83 8.75 LT0: 11.7 | LT1: 11.7 | LT2: 11.7 LT0: 11.7 | LT1: 11.7
9 RP 5.8 (Uniform thresh 17.5) 5.83 8.75 11.67 LT0: 17.5 | LT1: 17.5 | LT2: 17.5 LT0: 17.5 | LT1: 17.5

Trigger Activation Detail (per leadtime)

Click a circled point on the tradeoff plot above to swap the detailed per-leadtime activation table below.

Table 9.12
Trigger Activation: SLV (1991–2024)
Green = trigger fires at that leadtime | Annual RP = 2.92 (RP 2.9 A)
Year P CERF EM-DAT Affected Primera Forecast RP (3.9) Postrera Forecast RP (3.9) ERA5 Obs RP
Reported Interpolated LT0 (May) RP 5.0 LT1 (Apr) RP 5.0 LT2 (Mar) RP 5.0 LT0 (Sep) RP 5.0 LT1 (Aug) RP 5.0 Pri Post
1991
11.7 3.9 7.0 35.0 17.5 5.8 8.8
1992
3.2 2.7 1.3 5.8 4.4 2.1 5.0
1994 2
0
0
5.0 4.4 8.8 8.8 7.0 17.5 4.4
1997
7.0 5.8 3.5 4.4 35.0 11.7 2.1
2002
3.5 2.1 1.2 3.2 5.8 2.7 3.5
2009 1
0
0
8.8 11.7 35.0 17.5 5.0 2.3 5.8
2012
2.9 2.9 3.9 5.0 3.9 3.2 35.0
2014 1 3.9 8.8 11.7 3.5 3.5 3.9 1.8
2015 1 ✓
700K
350K
35.0 35.0 17.5 11.7 11.7 35.0 1.2
2018 1
387K
193K
17.5 7.0 5.0 7.0 8.8 8.8 1.5
2019 2 ✓
193K
5.8 17.5 5.8 1.8 2.1 2.9 1.6
2023
2.5 5.0 4.4 2.2 2.3 5.0 2.2
1993
1.5 1.8 2.3 2.7 2.9 1.5 1.9
1995
1.6 1.3 1.2 1.1 1.2 1.1 1.3
1996
1.2 1.2 1.3 1.9 2.2 1.2 1.8
1998
0
0
1.3 1.2 1.0 1.1 1.1 1.8 1.0
1999
1.2 1.2 1.5 1.2 1.1 1.6 1.1
2000
1.8 1.6 1.4 1.6 1.8 3.5 2.7
2001 2
400K
400K
2.1 2.3 2.1 3.9 2.7 7.0 11.7
2003
1.7 1.5 1.6 2.3 1.5 1.9 3.9
2004
1.9 1.4 1.8 2.5 3.2 2.2 2.3
2005
1.1 1.1 1.1 1.5 1.2 1.1 1.3
2006
1.8 1.5 2.7 2.1 1.7 1.8 2.5
2007
1.1 1.1 1.2 1.3 1.6 1.7 1.5
2008
1.5 1.8 1.7 1.8 1.2 1.2 1.1
2010
1.0 1.0 1.1 1.0 1.1 1.0 3.2
2011
1.2 1.7 2.5 1.2 1.0 1.3 1.2
2013
1.3 1.3 1.8 1.7 1.4 2.5 2.9
2016
350K
2.2 2.2 1.5 1.5 1.8 4.4 7.0
2017
2.3 3.5 3.2 1.4 1.9 1.4 1.7
2020
2.7 1.9 2.2 1.3 1.5 1.2 1.1
2021
4.4 3.2 2.9 2.9 2.5 1.5 17.5
2022
1.4 2.5 1.9 1.1 1.3 1.3 1.2
2024
1.1 1.1 1.1 1.2 1.3 1.1 1.4

Activation Summary

Table 9.13
El Salvador: Activation Summary
Pr = primera only | Pst = postrera only | Pr+Pst = both seasons
Year P CERF EM-DAT ERA5 Obs RP Activation
Pri Post RP 2.7 A RP 2.9 A RP 3.2 A RP 3.5 A RP 2.7 (Uniform thresh 4.4) RP 2.9 (Uniform thresh 5.0) RP 3.5 (Uniform thresh 5.8) RP 4.4 (Uniform thresh 7.0) RP 4.4 (Uniform thresh 8.8) RP 5.0 (Uniform thresh 11.7) RP 5.8 (Uniform thresh 17.5)
1991 5.8 8.8 Pr+Pst Pr+Pst Pr+Pst Pr+Pst Pr+Pst Pr+Pst Pr+Pst Pr+Pst Pr+Pst Pr+Pst Pst
1992 2.1 5.0 Pst Pst Pst Pst Pst Pst Pst
1994 P2
0
17.5 4.4 Pr+Pst Pr+Pst Pr+Pst Pr+Pst Pr+Pst Pr+Pst Pr+Pst Pr+Pst Pr+Pst
1997 11.7 2.1 Pr+Pst Pr+Pst Pr+Pst Pr+Pst Pr+Pst Pr+Pst Pr+Pst Pr+Pst Pst Pst Pst
2002 2.7 3.5 Pst Pst Pst Pst Pst
2009 P1
0
2.3 5.8 Pr+Pst Pr+Pst Pr+Pst Pr+Pst Pr+Pst Pr+Pst Pr+Pst Pr+Pst Pr+Pst Pr+Pst Pr+Pst
2012 3.2 35.0 Pst Pst Pst Pst Pst
2014 P1 3.9 1.8 Pr Pr Pr Pr Pr Pr Pr Pr Pr Pr
2015 P1 ✓
350K
35.0 1.2 Pr+Pst Pr+Pst Pr+Pst Pr+Pst Pr+Pst Pr+Pst Pr+Pst Pr+Pst Pr+Pst Pr+Pst Pr
2018 P1
193K
8.8 1.5 Pr+Pst Pr+Pst Pr+Pst Pr+Pst Pr+Pst Pr+Pst Pr+Pst Pr+Pst Pr+Pst Pr Pr
2019 P2 ✓
193K
2.9 1.6 Pr Pr Pr Pr Pr Pr Pr Pr Pr Pr Pr
2021 1.5 17.5 Pr Pr
2023 5.0 2.2 Pr Pr Pr Pr Pr Pr
1993 1.5 1.9
1995 1.1 1.3
1996 1.2 1.8
1998
0
1.8 1.0
1999 1.6 1.1
2000 3.5 2.7
2001 P2
400K
7.0 11.7
2003 1.9 3.9
2004 2.2 2.3
2005 1.1 1.3
2006 1.8 2.5
2007 1.7 1.5
2008 1.2 1.1
2010 1.0 3.2
2011 1.3 1.2
2013 2.5 2.9
2016
350K
4.4 7.0
2017 1.4 1.7
2020 1.2 1.1
2022 1.3 1.2
2024 1.1 1.4
Table 9.14: El Salvador recommended configurations: best average F1 per return period, tiebroken by lowest cross-season threshold CV.
El Salvador: Recommended Configurations
A = max avg F1 | B = max ERA5 tol F1 | U = uniform (all thresholds equal)
Return periods Return periods
config_label Postrera thresholds (RP) Primera thresholds (RP) annual_rp Performance
Criterion s0 s1 s_seasonal_rp p0 p1 p2 p_seasonal_rp impact_f1 era5_tol_f1 avg_f1 mean_cv all_cv
RP 2.7 A both 4.375000 4.375000 3.89 4.375000 4.375000 4.375000 3.50 2.69 0.513 0.697 0.605 0.000 0.000
RP 2.9 A both 5.000000 5.000000 3.89 5.000000 5.000000 5.000000 3.89 2.92 0.529 0.737 0.633 0.000 0.000
RP 3.2 A both 5.000000 7.000000 4.38 5.000000 5.000000 5.833333 3.89 3.18 0.544 0.734 0.639 0.163 0.158
RP 3.5 A both 5.833333 7.000000 5.00 5.000000 5.000000 5.833333 3.89 3.50 0.561 0.688 0.624 0.110 0.143
RP 2.7 (Uniform thresh 4.4) uniform 4.375000 4.375000 3.89 4.375000 4.375000 4.375000 3.50 2.69 0.513 0.697 0.605 0.000 0.000
RP 2.9 (Uniform thresh 5.0) uniform 5.000000 5.000000 3.89 5.000000 5.000000 5.000000 3.89 2.92 0.529 0.737 0.633 0.000 0.000
RP 3.5 (Uniform thresh 5.8) uniform 5.833333 5.833333 4.38 5.833333 5.833333 5.833333 4.38 3.50 0.562 0.647 0.605 0.000 0.000
RP 4.4 (Uniform thresh 7.0) uniform 7.000000 7.000000 5.83 7.000000 7.000000 7.000000 4.38 4.38 0.598 0.520 0.559 0.000 0.000
RP 4.4 (Uniform thresh 8.8) uniform 8.750000 8.750000 5.83 8.750000 8.750000 8.750000 5.00 4.38 0.619 0.452 0.536 0.000 0.000
RP 5.0 (Uniform thresh 11.7) uniform 11.666667 11.666667 8.75 11.666667 11.666667 11.666667 5.83 5.00 0.452 0.452 0.452 0.000 0.000
RP 5.8 (Uniform thresh 17.5) uniform 17.500000 17.500000 11.67 17.500000 17.500000 17.500000 8.75 5.83 0.424 0.424 0.424 0.000 0.000
Table 9.15
El Salvador: Topline Performance
34-year record (1991–2024) | 8 EM-DAT impact years
Config Annual RP Priority Years1 vs EM-DAT Impact2 vs ERA5 Rainfall3 Years Fired
P1 Caught P2 Caught TP FP FN TN F1 TP FP FN TN F1
RP 2.9 A 2.92 4/4 2/3 5 7 3 19 0.5 12 (14) 6 (4) 6 44 0.737 1991 1992 1994 1997 2002 2009 2012 2014 2015 2018 2019 2023
Primera Both Postrera Red outline = not all configs fire
1 P1 = CERF-funded drought years (highest confidence). P2 = all 3 countries in EM-DAT (regional consensus).
2 TP = trigger fires in an impact year. FP = fires in a non-impact year (may still be genuinely dry). FN = misses an impact year.
3 ERA5: summed across primera + postrera. TP & FP shown as matched (tolerant). FN = matched only (tolerant FN equals matched by design). Tolerant = widened drought definition (RP − buffer). F1 computed from tolerant TP, tolerant FP, matched FN.

Guatemala uses departments Chiquimula (GT20) and Alta Verapaz (GT16) as the area of interest, with 3 primera leadtimes and 2 postrera leadtimes. EM-DAT records 9 impact years in 1991–2024, giving an empirical impact RP of 3.0 years.

Impact Evaluation & Priority Filtering

After generating 2,772 constrained configurations and evaluating against impact years, priority filtering retains 28 candidates that hit 100% of Priority 1 years and 67% of Priority 2 years.

ERA5 Evaluation & Combined Scoring

Figure 9.3: ERA5 tolerant F1 vs Impact F1 for Guatemala. Click a circled point to see its trigger activation detail below.
Table 9.16
Guatemala: Configuration Summary
# Config Annual RP Primera RP Postrera RP Primera Thresholds Postrera Thresholds
1 RP 2.3 (Uniform thresh 4.4) 2.33 2.92 3.89 LT0: 4.4 | LT1: 4.4 | LT2: 4.4 LT0: 4.4 | LT1: 4.4
2 RP 2.7 A 2.69 3.50 3.89 LT0: 4.4 | LT1: 5.8 | LT2: 5.8 LT0: 4.4 | LT1: 4.4
3 RP 2.7 (Uniform thresh 5.0) 2.69 3.50 4.38 LT0: 5.0 | LT1: 5.0 | LT2: 5.0 LT0: 5.0 | LT1: 5.0
4 RP 2.9 A 2.92 3.89 3.89 LT0: 4.4 | LT1: 11.7 | LT2: 17.5 LT0: 4.4 | LT1: 4.4
5 RP 3.2 (Uniform thresh 5.8) 3.18 3.89 5.00 LT0: 5.8 | LT1: 5.8 | LT2: 5.8 LT0: 5.8 | LT1: 5.8
6 RP 3.2 (Uniform thresh 7.0) 3.18 3.89 5.83 LT0: 7.0 | LT1: 7.0 | LT2: 7.0 LT0: 7.0 | LT1: 7.0
7 RP 3.5 (Uniform thresh 8.8) 3.50 4.38 5.83 LT0: 8.8 | LT1: 8.8 | LT2: 8.8 LT0: 8.8 | LT1: 8.8
8 RP 3.5 (Uniform thresh 11.7) 3.50 5.00 5.83 LT0: 11.7 | LT1: 11.7 | LT2: 11.7 LT0: 11.7 | LT1: 11.7
9 RP 4.4 (Uniform thresh 17.5) 4.38 7.00 8.75 LT0: 17.5 | LT1: 17.5 | LT2: 17.5 LT0: 17.5 | LT1: 17.5

Trigger Activation Detail (per leadtime)

Click a circled point on the tradeoff plot above to swap the detailed per-leadtime activation table below.

Table 9.17
Trigger Activation: GTM (1991–2024)
Green = trigger fires at that leadtime | Annual RP = 2.92 (RP 2.9 A)
Year P CERF EM-DAT Affected Primera Forecast RP (3.9) Postrera Forecast RP (3.9) ERA5 Obs RP
Reported Interpolated LT0 (May) RP 4.4 LT1 (Apr) RP 11.7 LT2 (Mar) RP 17.5 LT0 (Sep) RP 4.4 LT1 (Aug) RP 4.4 Pri Post
1991
8.8 2.2 4.4 7.0 17.5 5.0 4.4
1992
1.5 1.6 1.1 11.7 5.8 1.5 3.9
1994 2
0
0
7.0 3.9 17.5 35.0 8.8 17.5 5.8
1997
2.7 1.8 1.6 4.4 35.0 2.9 1.5
2009 1 ✓
2.5M
2.5M
5.8 35.0 35.0 5.8 5.0 2.7 7.0
2012
266K
266K
4.4 3.2 3.2 5.0 3.5 1.3 17.5
2014 1 ✓
1.3M
433K
1.2 2.3 2.3 3.2 4.4 1.8 1.2
2015 1
433K
11.7 8.8 2.2 17.5 7.0 11.7 1.2
2018 1
1.5M
750K
17.5 5.8 7.0 8.8 11.7 7.0 5.0
2019 2 ✓
750K
5.0 7.0 3.5 2.2 3.2 35.0 2.2
2021
35.0 11.7 5.8 3.5 2.2 1.8 35.0
2023
2.2 17.5 8.8 2.5 1.8 8.8 1.6
1993
1.3 1.4 1.5 2.7 2.9 1.2 3.2
1995
2.1 1.5 1.2 1.1 1.1 1.4 1.8
1996
1.2 1.7 1.8 2.3 2.3 1.3 1.7
1998
1.7 1.2 1.0 1.1 1.2 5.8 1.1
1999
1.5 1.2 2.5 1.2 1.1 2.1 1.2
2000
1.9 3.5 1.9 1.8 1.9 1.5 2.1
2001 2
114K
114K
2.9 2.9 2.7 3.9 2.7 3.5 2.9
2002
3.9 2.1 1.3 1.9 3.9 4.4 3.5
2003
2.3 1.5 1.7 2.1 1.5 3.2 1.9
2004
1.2 1.1 1.4 1.8 2.5 2.3 11.7
2005
1.1 1.1 1.1 1.3 1.2 1.1 2.5
2006
3.2 1.8 11.7 2.9 1.5 1.1 2.3
2007
1.1 1.2 1.2 1.7 1.3 2.5 1.8
2008
2.5 4.4 2.1 1.3 1.2 1.7 1.1
2010
1.0 1.0 1.1 1.0 1.0 1.0 8.8
2011
1.1 2.7 3.9 1.1 1.1 1.9 1.1
2013
1.3 1.3 1.5 1.4 1.6 1.1 1.4
2016
✓
433K
1.4 1.3 1.3 1.6 2.1 3.9 2.7
2017
1.8 2.5 1.8 1.5 1.8 1.2 1.5
2020
1.8 1.9 2.9 1.2 1.4 1.2 1.0
2022
3.5 5.0 5.0 1.5 1.7 1.6 1.3
2024
1.6 1.1 1.2 1.2 1.3 2.2 1.3

Activation Summary

Table 9.18
Guatemala: Activation Summary
Pr = primera only | Pst = postrera only | Pr+Pst = both seasons
Year P CERF EM-DAT ERA5 Obs RP Activation
Pri Post RP 2.7 A RP 2.9 A RP 2.3 (Uniform thresh 4.4) RP 2.7 (Uniform thresh 5.0) RP 3.2 (Uniform thresh 5.8) RP 3.2 (Uniform thresh 7.0) RP 3.5 (Uniform thresh 8.8) RP 3.5 (Uniform thresh 11.7) RP 4.4 (Uniform thresh 17.5)
1991 5.0 4.4 Pr+Pst Pr+Pst Pr+Pst Pr+Pst Pr+Pst Pr+Pst Pr+Pst Pst Pst
1992 1.5 3.9 Pst Pst Pst Pst Pst Pst Pst Pst
1994 P2
0
17.5 5.8 Pr+Pst Pr+Pst Pr+Pst Pr+Pst Pr+Pst Pr+Pst Pr+Pst Pr+Pst Pr+Pst
1997 2.9 1.5 Pst Pst Pst Pst Pst Pst Pst Pst Pst
2006 1.1 2.3 Pr Pr Pr Pr Pr Pr Pr
2008 1.7 1.1 Pr
2009 P1 ✓
2.5M
2.7 7.0 Pr+Pst Pr+Pst Pr+Pst Pr+Pst Pr+Pst Pr Pr Pr Pr
2012
266K
1.3 17.5 Pr+Pst Pr+Pst Pr+Pst Pst
2014 P1 ✓
433K
1.8 1.2 Pst Pst Pst
2015 P1
433K
11.7 1.2 Pr+Pst Pr+Pst Pr+Pst Pr+Pst Pr+Pst Pr+Pst Pr+Pst Pr+Pst Pst
2018 P1
750K
7.0 5.0 Pr+Pst Pr+Pst Pr+Pst Pr+Pst Pr+Pst Pr+Pst Pr+Pst Pr+Pst Pr
2019 P2 ✓
750K
35.0 2.2 Pr Pr Pr Pr Pr Pr
2021 1.8 35.0 Pr Pr Pr Pr Pr Pr Pr Pr Pr
2022 1.6 1.3 Pr Pr
2023 8.8 1.6 Pr Pr Pr Pr Pr Pr Pr Pr Pr
1993 1.2 3.2
1995 1.4 1.8
1996 1.3 1.7
1998 5.8 1.1
1999 2.1 1.2
2000 1.5 2.1
2001 P2
114K
3.5 2.9
2002 4.4 3.5
2003 3.2 1.9
2004 2.3 11.7
2005 1.1 2.5
2007 2.5 1.8
2010 1.0 8.8
2011 1.9 1.1
2013 1.1 1.4
2016 ✓
433K
3.9 2.7
2017 1.2 1.5
2020 1.2 1.0
2024 2.2 1.3
Table 9.19: Guatemala recommended configurations: best average F1 per return period, tiebroken by lowest cross-season threshold CV.
Guatemala: Recommended Configurations
A = max avg F1 | B = max ERA5 tol F1 | U = uniform (all thresholds equal)
Return periods Return periods
config_label Postrera thresholds (RP) Primera thresholds (RP) annual_rp Performance
Criterion s0 s1 s_seasonal_rp p0 p1 p2 p_seasonal_rp impact_f1 era5_tol_f1 avg_f1 mean_cv all_cv
RP 2.7 A both 4.375000 4.375000 3.89 4.375000 5.833333 5.833333 3.50 2.69 0.649 0.667 0.658 0.079 0.161
RP 2.9 A both 4.375000 4.375000 3.89 4.375000 11.666667 17.500000 3.89 2.92 0.667 0.667 0.667 0.294 0.705
RP 2.3 (Uniform thresh 4.4) uniform 4.375000 4.375000 3.89 4.375000 4.375000 4.375000 2.92 2.33 0.619 0.613 0.616 0.000 0.000
RP 2.7 (Uniform thresh 5.0) uniform 5.000000 5.000000 4.38 5.000000 5.000000 5.000000 3.50 2.69 0.557 0.557 0.557 0.000 0.000
RP 3.2 (Uniform thresh 5.8) uniform 5.833333 5.833333 5.00 5.833333 5.833333 5.833333 3.89 3.18 0.528 0.528 0.528 0.000 0.000
RP 3.2 (Uniform thresh 7.0) uniform 7.000000 7.000000 5.83 7.000000 7.000000 7.000000 3.89 3.18 0.478 0.478 0.478 0.000 0.000
RP 3.5 (Uniform thresh 8.8) uniform 8.750000 8.750000 5.83 8.750000 8.750000 8.750000 4.38 3.50 0.435 0.435 0.435 0.000 0.000
RP 3.5 (Uniform thresh 11.7) uniform 11.666667 11.666667 5.83 11.666667 11.666667 11.666667 5.00 3.50 0.450 0.450 0.450 0.000 0.000
RP 4.4 (Uniform thresh 17.5) uniform 17.500000 17.500000 8.75 17.500000 17.500000 17.500000 7.00 4.38 0.368 0.368 0.368 0.000 0.000

The StartNetwork Guatemala AOI uses departments Quiché (GT14) and Baja Verapaz (GT15), with 3 primera leadtimes and 2 postrera leadtimes. Impact years are drawn from the same country-level EM-DAT record (9 events, empirical RP = 3.0).

Impact Evaluation & Priority Filtering

After priority filtering, 84 candidates remain, hitting 100% of Priority 1 years and 100% of Priority 2 years.

ERA5 Evaluation & Combined Scoring

Figure 9.4: ERA5 tolerant F1 vs Impact F1 for Guatemala (StartNetwork). Click a circled point to see its trigger activation detail below.
Table 9.20
Guatemala (StartNetwork): Configuration Summary
# Config Annual RP Primera RP Postrera RP Primera Thresholds Postrera Thresholds
1 RP 2.7 A 2.69 4.38 3.18 LT0: 4.4 | LT1: 5.8 | LT2: 7.0 LT0: 4.4 | LT1: 5.0
2 RP 2.9 A 2.92 4.38 3.50 LT0: 4.4 | LT1: 5.8 | LT2: 7.0 LT0: 4.4 | LT1: 5.8
3 RP 2.9 (Uniform thresh 5.8) 2.92 4.38 4.38 LT0: 5.8 | LT1: 5.8 | LT2: 5.8 LT0: 5.8 | LT1: 5.8
4 RP 2.2 (Uniform thresh 4.4) 2.19 3.18 3.18 LT0: 4.4 | LT1: 4.4 | LT2: 4.4 LT0: 4.4 | LT1: 4.4
5 RP 2.5 (Uniform thresh 5.0) 2.50 3.89 3.50 LT0: 5.0 | LT1: 5.0 | LT2: 5.0 LT0: 5.0 | LT1: 5.0
6 RP 3.5 (Uniform thresh 7.0) 3.50 5.00 5.83 LT0: 7.0 | LT1: 7.0 | LT2: 7.0 LT0: 7.0 | LT1: 7.0
7 RP 3.9 (Uniform thresh 8.8) 3.89 5.00 7.00 LT0: 8.8 | LT1: 8.8 | LT2: 8.8 LT0: 8.8 | LT1: 8.8
8 RP 4.4 (Uniform thresh 11.7) 4.38 7.00 7.00 LT0: 11.7 | LT1: 11.7 | LT2: 11.7 LT0: 11.7 | LT1: 11.7
9 RP 5.0 (Uniform thresh 17.5) 5.00 8.75 8.75 LT0: 17.5 | LT1: 17.5 | LT2: 17.5 LT0: 17.5 | LT1: 17.5

Trigger Activation Detail (per leadtime)

Click a circled point on the tradeoff plot above to swap the detailed per-leadtime activation table below.

Table 9.21
Trigger Activation: GTM (1991–2024)
Green = trigger fires at that leadtime | Annual RP = 2.92 (RP 2.9 A)
Year P CERF EM-DAT Affected Primera Forecast RP (4.4) Postrera Forecast RP (3.5) ERA5 Obs RP
Reported Interpolated LT0 (May) RP 4.4 LT1 (Apr) RP 5.8 LT2 (Mar) RP 7.0 LT0 (Sep) RP 4.4 LT1 (Aug) RP 5.8 Pri Post
1991
8.8 2.9 7.0 11.7 35.0 5.0 8.8
1992
1.8 2.1 1.2 35.0 8.8 1.8 3.2
1994 2
0
0
7.0 7.0 17.5 17.5 7.0 35.0 7.0
1997
3.5 2.7 1.8 3.5 17.5 2.7 1.3
2001 2
114K
114K
2.9 2.2 2.2 4.4 2.9 2.3 2.3
2009 1 ✓
2.5M
2.5M
5.0 35.0 35.0 5.0 3.5 2.5 11.7
2012
266K
266K
4.4 3.2 3.5 7.0 3.9 1.1 17.5
2014 1 ✓
1.3M
433K
1.2 3.9 2.9 2.3 5.8 2.1 1.2
2015 1
433K
35.0 17.5 5.0 5.8 4.4 8.8 1.5
2018 1
1.5M
750K
11.7 11.7 11.7 8.8 11.7 4.4 4.4
2019 2 ✓
750K
5.8 8.8 3.2 1.8 2.7 17.5 1.9
2021
17.5 5.8 8.8 3.9 2.3 1.5 35.0
1993
1.3 1.3 1.7 3.2 3.2 1.3 3.5
1995
1.9 1.5 1.2 1.1 1.1 1.5 1.4
1996
1.1 1.7 1.8 2.9 2.1 1.3 2.2
1998
1.5 1.2 1.1 1.1 1.2 2.9 1.0
1999
1.3 1.2 2.3 1.3 1.0 2.2 1.3
2000
1.5 2.3 1.5 1.8 1.9 1.6 1.8
2002
3.9 1.9 1.3 1.9 5.0 7.0 2.5
2003
2.5 1.4 1.6 2.2 1.5 3.9 2.7
2004
1.6 1.2 1.5 2.5 2.5 11.7 5.0
2005
1.1 1.1 1.1 1.4 1.2 1.2 2.1
2006
2.3 1.5 5.8 2.7 1.3 1.1 2.9
2007
1.2 1.1 1.2 1.7 1.4 3.2 1.6
2008
1.8 2.5 1.9 1.5 1.2 1.4 1.1
2010
1.0 1.0 1.0 1.0 1.1 1.0 3.9
2011
1.1 1.8 2.5 1.1 1.1 1.7 1.1
2013
1.2 1.3 1.4 1.2 1.7 1.1 1.2
2016
✓
433K
2.2 1.6 1.3 1.6 2.2 3.5 5.8
2017
2.7 3.5 2.1 1.5 1.8 1.2 1.8
2020
2.1 1.8 2.7 1.2 1.5 1.2 1.1
2022
3.2 4.4 3.9 1.3 1.8 1.9 1.2
2023
1.7 5.0 4.4 2.1 1.6 5.8 1.7
2024
1.4 1.1 1.1 1.2 1.3 1.8 1.5

Activation Summary

Table 9.22
Guatemala (StartNetwork): Activation Summary
Pr = primera only | Pst = postrera only | Pr+Pst = both seasons
Year P CERF EM-DAT ERA5 Obs RP Activation
Pri Post RP 2.7 A RP 2.9 A RP 2.2 (Uniform thresh 4.4) RP 2.5 (Uniform thresh 5.0) RP 2.9 (Uniform thresh 5.8) RP 3.5 (Uniform thresh 7.0) RP 3.9 (Uniform thresh 8.8) RP 4.4 (Uniform thresh 11.7) RP 5.0 (Uniform thresh 17.5)
1991 5.0 8.8 Pr+Pst Pr+Pst Pr+Pst Pr+Pst Pr+Pst Pr+Pst Pr+Pst Pst Pst
1992 1.8 3.2 Pst Pst Pst Pst Pst Pst Pst Pst Pst
1994 P2
0
35.0 7.0 Pr+Pst Pr+Pst Pr+Pst Pr+Pst Pr+Pst Pr+Pst Pr+Pst Pr+Pst Pr+Pst
1997 2.7 1.3 Pst Pst Pst Pst Pst Pst Pst Pst Pst
2001 P2
114K
2.3 2.3 Pst Pst Pst
2002 7.0 2.5 Pst Pst Pst
2006 1.1 2.9 Pr Pr Pr
2009 P1 ✓
2.5M
2.5 11.7 Pr+Pst Pr+Pst Pr+Pst Pr+Pst Pr Pr Pr Pr Pr
2012
266K
1.1 17.5 Pr+Pst Pr+Pst Pr+Pst Pst Pst Pst
2014 P1 ✓
433K
2.1 1.2 Pst Pst Pst Pst Pst
2015 P1
433K
8.8 1.5 Pr+Pst Pr+Pst Pr+Pst Pr+Pst Pr+Pst Pr Pr Pr Pr
2018 P1
750K
4.4 4.4 Pr+Pst Pr+Pst Pr+Pst Pr+Pst Pr+Pst Pr+Pst Pr+Pst Pr+Pst
2019 P2 ✓
750K
17.5 1.9 Pr Pr Pr Pr Pr Pr Pr
2021 1.5 35.0 Pr Pr Pr Pr Pr Pr Pr Pr Pr
2022 1.9 1.2 Pr
2023 5.8 1.7 Pr Pr
1993 1.3 3.5
1995 1.5 1.4
1996 1.3 2.2
1998 2.9 1.0
1999 2.2 1.3
2000 1.6 1.8
2003 3.9 2.7
2004 11.7 5.0
2005 1.2 2.1
2007 3.2 1.6
2008 1.4 1.1
2010 1.0 3.9
2011 1.7 1.1
2013 1.1 1.2
2016 ✓
433K
3.5 5.8
2017 1.2 1.8
2020 1.2 1.1
2024 1.8 1.5
Table 9.23: Guatemala (StartNetwork) recommended configurations: best average F1 per return period, tiebroken by lowest cross-season threshold CV.
Guatemala (StartNetwork): Recommended Configurations
A = max avg F1 | B = max ERA5 tol F1 | U = uniform (all thresholds equal)
Return periods Return periods
config_label Postrera thresholds (RP) Primera thresholds (RP) annual_rp Performance
Criterion s0 s1 s_seasonal_rp p0 p1 p2 p_seasonal_rp impact_f1 era5_tol_f1 avg_f1 mean_cv all_cv
RP 2.7 A both 4.375000 5.000000 3.18 4.375000 5.833333 7.000000 4.38 2.69 0.703 0.646 0.674 0.162 0.210
RP 2.9 A both 4.375000 5.833333 3.50 4.375000 5.833333 7.000000 4.38 2.92 0.721 0.598 0.660 0.216 0.204
RP 2.2 (Uniform thresh 4.4) uniform 4.375000 4.375000 3.18 4.375000 4.375000 4.375000 3.18 2.19 0.650 0.638 0.644 0.000 0.000
RP 2.5 (Uniform thresh 5.0) uniform 5.000000 5.000000 3.50 5.000000 5.000000 5.000000 3.89 2.50 0.594 0.594 0.594 0.000 0.000
RP 2.9 (Uniform thresh 5.8) uniform 5.833333 5.833333 4.38 5.833333 5.833333 5.833333 4.38 2.92 0.588 0.588 0.588 0.000 0.000
RP 3.5 (Uniform thresh 7.0) uniform 7.000000 7.000000 5.83 7.000000 7.000000 7.000000 5.00 3.50 0.513 0.513 0.513 0.000 0.000
RP 3.9 (Uniform thresh 8.8) uniform 8.750000 8.750000 7.00 8.750000 8.750000 8.750000 5.00 3.89 0.455 0.455 0.455 0.000 0.000
RP 4.4 (Uniform thresh 11.7) uniform 11.666667 11.666667 7.00 11.666667 11.666667 11.666667 7.00 4.38 0.429 0.429 0.429 0.000 0.000
RP 5.0 (Uniform thresh 17.5) uniform 17.500000 17.500000 8.75 17.500000 17.500000 17.500000 8.75 5.00 0.308 0.308 0.308 0.000 0.000

9.7 Appendix: ROC Analysis

The ROC-style analysis below examines how well ERA5 observed rainfall (at various RP thresholds) predicts EM-DAT drought events. Low AUC values confirm the weak relationship between rainfall severity ranking and humanitarian impact, further supporting the hybrid optimization approach used in this chapter.

Figure 9.5: ROC-style curve for Honduras: ERA5 rainfall RP threshold vs EM-DAT drought detection.
Figure 9.6: ROC-style curve for El Salvador: ERA5 rainfall RP threshold vs EM-DAT drought detection.
Figure 9.7: ROC-style curve for Guatemala: ERA5 rainfall RP threshold vs EM-DAT drought detection.
Figure 9.8: ROC-style curve for Guatemala (StartNetwork): ERA5 rainfall RP threshold vs EM-DAT drought detection.